Past and future climate variations in the Norwegian Arctic: overview and novel analyses

Eirik J. Førland & Inger Hanssen-Bauer

Sparse stations and serious measuring problems hamper analyses of climatic conditions in the Arctic. This paper presents a discussion of measuring problems in the Arctic and gives an overview of observed past and projected future climate variations in and Jan Mayen. Novel analyses of temperature conditions during precipitation and trends in fractions of solid/liquid precipitation at the Arctic weather stations are also outlined. Analyses based on combined and homogenized series from the regular weather stations in the region indicate that the measured annual precipitation has increased by more than 2.5 % per decade since the measurements started in the beginning of the 20th century. The annual temperature has increased in during the latest decades, but the present level is still lower than in the 1930s. Downscaled scenarios for Svalbard Airport indicate a further increase in temperature and precipitation. Analyses based on observations of precipitation types at the regular weather stations demonstrate that the annual fraction of solid precipitation has decreased at all stations during the latest decades. The reduced fraction of solid precipitation implies that the undercatch of the precipitation gauges is reduced. Consequently, part of the observed increase in the annual precipitation is fi ctitious and is due to a larger part of the “true” precipitation being caught by the gauges. With continued warming in the region, this virtual increase will be measured in addition to an eventual real increase.

E. J. Førland & I. Hanssen-Bauer, Norwegian Meteorological Institute, Box 43 Blindern, NO-0313 Oslo, , [email protected].

A combination of instrumental records and increase of temperature and precipitation in reconstructions from proxy sources indicate that high northern latitudes as greenhouse gas con- Arctic air temperatures in the 20th century were centrations increase (IPCC 2001). The warm- the highest in the past 400 years (Serreze et al. ing in high northern latitudes may be amplifi ed 2000). During the 20th century an increase in by feedbacks from changes in snow and sea ice annual precipitation has been observed at higher extent and thawing of permafrost. The freshwater northern latitudes (Hulme 1995; Dai et al. 1997). budget in the Arctic has become an increasingly The Norwegian Arctic has experienced a substan- important consideration in the context of global tial increase in precipitation during the 20th cen- climate change (Walsh et al. 1998) as it may be tury, but although the temperature has increased linked to the intermittency of North Atlantic deep during the latest decades there is no signifi cant water formation and the global thermohaline cir- long-term trend in annual temperatures (Førland culation, which is a major determinant of global et al. 2002; Hanssen-Bauer 2002). climate (Aagaard & Carmack 1989; Mysak et al. Global climate models project signifi cant 1990). The Arctic freshwater budget is driven by

Førland & Hanssen-Bauer 2003: Polar Research 22(2), 113–124 113 Fig.1. Location of current manual weather stations in the Norwegian Arctic. Bjørnøya: elevation 16 m a.s.l., start of measurements 1920. : 6 m a.s.l., 1944. Hornsund (Polish): 10 m a.s.l., 1978. Sveagruva 9 m a.s.l., 1978. (Russian): 70 m a.s.l., 1933. Svalbard Airport: 28 m a.s.l., 1975. Ny-Ålesund: 8 m a.s.l., 1974. Jan Mayen: 10 m a.s.l., 1921.

river runoff, accumulation/ablation of glaciers the precipitation trend was almost doubled after and precipitation over the Arctic Ocean. The adjusting the series. Accordingly, it is of crucial observed and projected increases in temperature importance to adjust series for inhomogeneities and precipitation in Svalbard and Jan Mayen (Fig. when they are used in studies of long-term cli- 1) thus have broad implications for Arctic and, mate variations. perhaps, global climate and the monitoring of The climate and long-term climatic variations climatic trends in these regions is therefore also at the Norwegian Arctic stations are described in important in a global context. a number of old and new publications, e.g. Bir- Analyses of climatic trends in the Arctic are keland (1930), Hesselberg & Johannessen (1958), hampered by the sparse station network and seri- Steffensen (1969, 1982), Hisdal (1976), Vinje ous measuring problems. The large year-to-year (1982), Hanssen-Bauer et al. (1990), Førland et al. variations imply that the climate signals have to (1997a) and Winther et al. (2003). Serreze et al. be strong to be statistically signifi cant. Although (2000) review observational evidence of recent the scenarios indicate substantial climate chang- changes in the Arctic environment. es in the Arctic, it is not necessarily the case that In this paper a discussion of measuring prob- the fi rst signifi cant “greenhouse signal” will lems in the Arctic and an overview of past and be found in this region. Real climatic trends future climate variations are given in the fi rst may be masked or amplifi ed when analyses are sections. In the last sections novel analyses of based upon inhomogeneous series. Studies have temperature conditions during precipitation and revealed that inhomogeneities in Arctic climate trends in fractions of solid/liquid precipitation at series are often of the same magnitude as typi- the Arctic stations are outlined, and it is demon- cal long-term trends (Hanssen-Bauer & Førland strated that the observed trends in measured pre- 1994; Nordli et al. 1996). For Arctic stations in cipitation may deviate from the real precipitation Canada, Mekis & Hogg (1999) demonstrated that trends.

114 Past and future climate variations in the Norwegian Arctic Meteorological measuring problems In the Arctic, typical values of the combined ∆ ∆ in the Arctic effect of wetting and evaporation loss ( PW + PE) for the Norwegian gauge is 0.10 - 0.15 mm/case for solid, liquid and mixed precipitation (Før- General measuring problems land & Hanssen-Bauer 2000). At most measur- ing sites, wind speed is the most important envi- The Arctic climate poses several serious chal- ronmental factor contributing to the undercatch lenges for the monitoring of the main weather of the precipitation gauges. Based on results elements. Icing and wet snow may cause mal- from the World Meteorological Organization functioning of the various sensors at the weather Solid Precipitation Measurement Intercompar- stations. During the polar night, the combination ison (Goodison et al. 1998), Nordic precipita- of darkness and harsh weather occasionally com- tion gauge studies (Førland et al. 1996) and fi eld plicates manual observations. measurements in Ny-Ålesund, Førland & Hans- The combination of dry snow, high wind speed sen-Bauer (2000) deduced correction models for and open tundra increase dramatically the meas- hourly and daily measurements in the Norwegian uring errors for precipitation and snow depth at precipitation gauge under Arctic conditions. The most stations in the Norwegian Arctic. Because correction factor for solid precipitation was found of blowing snow, the snow layer on the ground to increase with increasing wind speeds, and is likely to show large local variations (Winther decrease with increasing temperature. et al. 1998; Winther et al. 2003), and the regu- By considering typical values of wind speed, lar snow depth measurements at weather stations temperature and precipitation intensity, Førland are therefore seldom representative for the snow & Hanssen-Bauer (2000) suggested the following accumulation in the station area. Consequent- rough correction factors for unsheltered stations ly, measurements of snow cover and snow depth on : liquid precipitation kl = 1.15, solid have been given little priority at the Norwegian ks = 1.85 and mixed km = (0.5 * kl + 0.5 * ks) = 1.5. It Arctic weather stations. should be stressed that using these typical correc- tion factors does not necessarily produce accurate Errors in precipitation measurements estimates of true precipitation at particular indi- vidual stations. Further, the corrections do not Several types of errors are connected to precipita- refl ect differences in wind exposure between dif- tion measurements (Goodison et al. 1998). For the ferent measuring sites. , Førland et al. (1996) concluded that the real amount of precipitation (“true pre- Adjustment for inhomogeneities cipitation”) might be expressed as: Inhomogeneities in climatic series may be ∆ ∆ caused by relocation of sensors, changed envi- PC = k ⋅(Pm + PW + PE ), (1) ronment (such as new buildings) and instrumen- where PC is true precipitation, k is the correction tal improvements. To acquire reliable long-term factor due to aerodynamic effects, Pm is measured climate series in the Arctic is particularly com- ∆ precipitation, PW is precipitation lost by wetting, plicated. Because of the harsh weather condi- ∆ and PE is precipitation lost by evaporation from tions, even small changes at measuring sites may the gauge. Other error types, such as splash in/ cause substantial changes in measuring condi- out, instrumental errors, misreadings etc., are tions for precipitation, for example. Identifi ca- either corrected in routine quality controls or may tion of inhomogeneities in Arctic climate series is be neglected under Nordic climate conditions. further hampered by the sparse station network. However, drifting or blowing snow occasionally Homogeneity tests based upon comparison with cause substantial problems in the Arctic. “Precip- neighbouring stations are diffi cult to perform for itation” caused solely by blowing snow is exclud- Bjørnøya (the southernmost island of Svalbard) ed through the quality control at the Norwegian and Jan Mayen, for example, as the nearest neigh- Meteorological Institute, but there is often a com- bouring stations are more than 300 km away. A bination of precipitation and blowing snow. In detailed survey of the results of the homogeneity such cases it is diffi cult to distinguish the propor- analyses for Norwegian Arctic series is given by tions of real precipitation and blowing snow. Nordli et al. (1996).

Førland & Hanssen-Bauer 2003: Polar Research 22(2), 113–124 115 Fig. 2. Low-pass fi ltered series of annual temperature at Norwegian Arctic stations. The series are smoothed using Gaussian weight- ing coeffi cients and show vari- ations on a decadal time scale. (The last three years of the curves are omitted as fi ltered series are unreliable at the ends.)

Temperature and precipitation varia- midwinter, temperatures well above zero have tions in the 20th century been recorded at all stations (e.g. +12.3 °C at Jan Mayen on December 2001). During summer, Climate series for the Norwegian Arctic maximum temperatures above 20 °C have occa- sionally been recorded at Bjørnøya and Svalbard The available climate data from the Norwegian Airport. Arctic is rather limited. The present network of There are no signifi cant trends in the annual synoptical weather stations consists of fi ve sta- temperature at Svalbard Airport, Bjørnøya and tions on Spitsbergen and three stations on other Jan Mayen from the start of the series to the Arctic islands (Fig. 1). The oldest meteorologi- present (Table 1). However, a closer examination cal observations from the Norwegian Arctic were of the series reveals three sub-periods with sig- made during scientifi c expeditions to Svalbard nifi cant trends. From the start in the 1910s there and Jan Mayen. In 1911 a permanent weather sta- is a positive trend up to the late 1930s, a temper- tion was established in Green Harbour, western ature decrease from the 1930s to the 1960s, and Spitsbergen. Around 1920, weather stations were from the 1960s to the present the temperature established at Bjørnøya and Jan Mayen. has increased signifi cantly. Figure 2 shows that Long-term climate series for Spitsbergen despite the warming in the recent decades, the are established by joining series from different warmest two decades on an annual basis are still sites, and applying statistical methods (Nordli et al. 1996). For Svalbard Airport the combined series is mainly based upon observations from Table 1. Linear trends (°C per decade) in observed tem- Green Harbour/Barentsburg and / peratures. Statistical signifi cant trends (Mann-Kendall) are in boldface (5 % level) and underlined (1 % level). Svalbard Airport. The Ny-Ålesund series is based upon measurements from and Ny- 1912– 1910– 1946– 1976– Station Ålesund. The weather stations in Ny-Ålesund 2001 1945 1975 2001 and Svalbard Airport were both moved to their Bjørnøya (–0.01) (+0.08) –0.29 +0.49 present sites in 1975. The long-term temperature Hopen (+0.05) – –0.53 +0.84 and precipitation series are adjusted before 1975 Sveagruva – – – +1.84 b to be valid for the present sites. Svalbard Airport a +0.15 +1.20 –0.48 +0.78 Ny-Ålesund a –––0.40+0.42 Temperature Jan Mayen (–0.08) (–0.20) –0.71 +0.49 Northern Hemisphere +0.07 c +0.14 d –0.04 +0.31 e There are pronounced fl uctuations in Arctic cli- (land) mate on daily, monthly and annual time scales. Global (land) +0.06 c +0.11 d –0.01 +0.22 e The lowest recorded temperature at the Norwe- a Combined series before 1975 c 1901–2000 e 1976–2000 gian Arctic stations is –46.3 °C, but even during b 1979–2001 d 1901–1945

116 Past and future climate variations in the Norwegian Arctic the 1930s and the 1950s. An interesting feature Precipitation is that the long-term series from the Norwegian Arctic stations show that on a decadal time scale, The severe measuring problems discussed above local temperature minima and maxima largely create serious uncertainties in Arctic precipi- occur within the same decades for all seasons tation values. Precipitation in the Arctic is low (Førland et al. 1997a). because air masses are usually stably strati- Because of substantial seasonal differences in fi ed and contain only small amounts of water standard deviations, the variation in annual mean vapour. The normal (1961–1990) annual precipi- temperatures is generally more affected by the tation at Svalbard Airport (190 mm) is the lowest variation in winter temperature than by summer at any Norwegian station. There are large local temperatures. Accordingly, both for the warming gradients in precipitation between the Spitsber- up to the 1930s and for the decrease in annual tem- gen stations: although the distance is just 35 km, perature from the 1930s to the 1960s, variations the annual precipitation at Barentsburg is almost in winter temperatures were dominant. However, three times as high as at Svalbard Airport. By during the latest three decades, increased spring measuring several transects, Sand et al. (2003) temperatures gave the largest single contribution revealed both west–east and south–north gradi- to the increase in annual temperature. ents of snow accumulation on Spitsbergen. They The temperature level at the Norwegian Arctic found that the winter accumulation rates along stations is currently somewhat lower than in the the east coast were about 40 % higher than on the 1930s. This is contrary to the rest of northern west coast, and that the southern part of the island Europe and for the globe as a whole, where the receives about twice as much winter precipitation present level is signifi cantly higher than the 1930s as the northern part. Large local precipitation level (Parker & Alexander 2002). gradients were also found for the Ny-Ålesund In Table 1 temperature trends are given for the area (Førland et al. 1997b). The precipitation dis- sub-periods applied in the latest Intergovernmen- tribution was found to be strongly dependant on tal Panel on Climate Change report (Folland & the large-scale wind direction. With winds from Karl 2001). During 1946–1975, all stations expe- the south and south-west, the precipitation at a rienced a negative trend of –0.3 to –0.7 °C/decade. glacier (Brøggerbreen) a few kilometres south- For recent decades (1976–2001), the trend is posi- west of Ny-Ålesund was about 60 % higher than tive at all stations, with the strongest warming at in Ny-Ålesund, while with winds from the north- Hopen, Sveagruva and Svalbard Airport. Despite west Ny-Ålesund received more precipitation the rather strong trends since 1976, the Svalbard than the glacier. Airport series is the only one with a statistically Annual precipitation (Fig. 3) has increased sub- signifi cant trend at the 1 % level. stantially during the 20th century at most of the Hanssen-Bauer & Førland (1998a) found that stations in the Norwegian Arctic. Both at Sval- though the temperature increase in the Norwe- bard Airport and Bjørnøya the increase is larger gian Arctic during the latest decades to a large than 2.5 % per decade (Table 2). At Jan Mayen degree may be explained by changes in atmos- most of the increase happened before 1960, while pheric circulation, this is apparently not the case the increase at the other stations is more evenly with respect to the early 20th century warming. distributed throughout the 20th century. A model based on regional atmospheric circula- The course of the precipitation increase at tion indices was able to account for most of the Svalbard parallels the increase in coastal parts trend from the 1960s to the present, but only of northern Norway (Hanssen-Bauer & Førland about one-third of the observed temperature 1998b), with a trend that seems to be fairly con- increase at Svalbard from 1912 to the 1930s and stant throughout the 20th century. However, the for the subsequent temperature decrease from the relative precipitation increase on Spitsbergen and 1930s to the 1960s. Fu et al. (1999) suggest that Bjørnøya is considerably higher than the con- ocean circulation and sea surface temperatures current increases on the Norwegian mainland. may be important for explaining the warming in It is also higher than the “average high latitude the northern North Atlantic region before 1940; increase” estimated by Hulme (1995). they conclude that this warming is not yet fully The observed long-term variations in precipi- understood. tation on the west coast of Spitsbergen during the 20th century may be explained largely by varia-

Førland & Hanssen-Bauer 2003: Polar Research 22(2), 113–124 117 Fig. 3. Low-pass fi ltered series of measured annual precipitation at Norwegian Arctic stations. The series are smoothed using Gaus- sian weighting coeffi cients and show variations on a decadal time scale. (The last three years of the curves are omitted as fi ltered series are unreliable at the ends.)

tions in the average atmospheric circulation con- times greater than that observed for the last 90 ditions (Hanssen-Bauer & Førland 1998a). Hans- years (Table 3). A similar warming rate is project- sen-Bauer (2002) concluded that about 70 % of ed in this area by dynamical downscaling based the trend in annual precipitation at Svalbard Air- upon the same climate model (Bjørge et al. 2000). port during the period 1912–1997 was accounted The projected warming is statistically signifi cant for by variations in atmospheric circulation. at the 1 % level in all seasons. Hanssen-Bauer & Førland (2001) conclud- ed that less than 20 % of the warming projected Future climate development by the GSDIO integration in Svalbard could be accounted for by changes in the atmospheric cir- Scenarios of temperature and precipitation for culation. When compared to the warming during Svalbard for the next 50 years have been worked recent decades, this implies that a diminishing out by Hanssen-Bauer (2002) by empirical down- part of the projected warming will be attributed scaling of an integration with the Max-Planck to changes in circulation. Some of the warming Institute’s coupled atmosphere–ocean global cli- is probably directly connected to the greenhouse mate model ECHAM4/OPYC3. The integration warming. However, the GSDIO integration shows (GSDIO) has a physical parameterization which extensive melting of sea ice east of Svalbard, and accounts for direct and indirect effects of sulphur feedback effects from the melting probably con- aerosols in addition to greenhouse gases, includ- tribute signifi cantly to the strong warming in the ing tropospheric ozone (Roeckner et al. 1999), area (Bjørge et al. 2000). Benestad et al. (2002) and is based upon the emission scenario IS92a compared downscaled temperature scenarios for (Houghton et al. 1992). Svalbard based upon three different global cli- The trend in the downscaled temperatures mate models. They concluded that differences from the GSDIO integration for the period 1960– between the models concerning sea ice conditions 2000 is mainly in accordance with what has been lead to highly variable local temperature projec- observed during that period, while the project- tions in Svalbard. The realism of future tempera- ed annual warming rate up to 2050 is almost fi ve ture scenarios is therefore critically dependent on

Table 2. Linear trends (% per decade) in observed and projected annual and seasonal precipitation. Statistical signifi cant trends (Mann-Kendall) are in boldface (5 % level) and underlined (1 % level). The magnitude of the trend is relative to the (observed) 1961–1990 normal.

Station Period Annual Winter Spring SummerAutumn Bjørnøya 1920–2001 +2.8 +3.9 +4.8 +1.3 +2.1 Jan Mayen 1921–2001 +1.7 +2.5 +4.6 +1.9 –0.6 Svalbard Airport (observed) 1912–2001 +2.7 –0.2 +2.2 +4.9 +3.7 Svalbard Airport (projected) 1961–2050 +1.4 +1.7 +4.6 -0.9 +0.6

118 Past and future climate variations in the Norwegian Arctic the reliability of the projected changes in the sea bard Airport and Ny-Ålesund (Table 4). Based ice concentrations in the region. on four observations per day (00, 06, 12 and 18 The downscaled precipitation scenario (Hans- UTC), the average number of events at Jan Mayen sen-Bauer 2002) also indicates that annual pre- implies that there is precipitation about 31 % of cipitation will increase signifi cantly up to 2050, the time. The value for Svalbard Airport is 20 %. mainly because of a highly signifi cant project- Snow is the most frequent precipitation type. ed increase in spring precipitation (Table 2, At Svalbard Airport, for example, around 75 % bottom row). Dynamical downscaling (Bjørge of precipitation events are reported as snow. At et al. 2000) projects an even higher precipitation all stations, 15 - 20 % of precipitation events are increase (ca. 2 % per decade) at the west coast of reported as rain and about 5 % as sleet. Drizzle Spitsbergen. constitutes 20 % of precipitation events at Jan Analyses of the Svalbard Airport series indi- Mayen but just 4 % at Svalbard Airport. cate that precipitation variation to a larger degree than temperature variation may be explained by Air temperature during precipitation changes in the atmospheric circulation. This is at least partly due to the local topography, which At Svalbard Airport, 35 % of snow events are shelters against precipitation from some sectors reported at air temperatures below –10 °C; at while it orographically enhances the precipitation Jan Mayen the value is 13 % (Fig. 4a). At all sta- from other sectors. This is also valid for the cli- tions less than 10 % of snow events are observed mate scenario, but the infl uence of atmospheric at temperatures above 0 °C. At Svalbard Airport, circulation on the long-term trend in the scenar- 30 % of the cases with sleet are reported for tem- io is considerably smaller than it has been during peratures > + 2°C; at Bjørnøya just 5 % (Fig. 4b). the 20th century. A major part of the projected For rain (Fig. 4c), around 25 % of the events at precipitation trend is accounted for by the tem- Svalbard Airport are reported at temperatures perature increase, which is used in the empiri- lower than 3 °C; at Bjørnøya the percentage is cal downscaling models as a proxy for increased above 50 %. The geographical differences in tem- air humidity. As different climate models seem to perature distribution are smaller for drizzle (Fig. show a closer agreement concerning the temper- 4d) than for rain. At Svalbard Airport and Ny- ature signal than concerning changes in atmos- Ålesund, ca. 30 % of drizzle events are reported pheric circulation (Räisänen 2001), this part of at temperatures below +3 °C; at Jan Mayen and the trend is probably also more credible than the Bjørnøya around 40 %. part caused by variations in the atmospheric cir- Wildlife, particularly the reindeer on Spits- culation. bergen, are vulnerable to events with snow crust or icy conditions. Extensive starving and death Precipitation characteristics Table 4. Frequencies of various precipitation types. The fre- Frequencies of different precipitation types quencies are given as percentage of the total number of cases based on 4 observations per day (at 00, 06, 12 and 18 UTC). The frequency of precipitation events is substan- However, at Ny-Ålesund there are no observations at 00 UTC and the number of cases is adjusted to 4 per day. tially higher at the island stations Bjørnøya and Jan Mayen than at the Spitsbergen stations Sval- Svalbard Ny- Jan Bjørnøya Airport Ålesund Mayen No. of cases Table 3. Linear trends (°C per decade) in observed and pro- 365.2 289.0 295.6 455.9 per year jected annual and seasonal temperature at Svalbard Airport. Statistical signifi cant trends (Mann-Kendall) are underlined Drizzle (%) 16.5 3.9 8.6 20.1 (1 % level). (From Hanssen-Bauer 2002.) Rain (%) 19.9 16.5 17.0 19.5 Sleet (%) 5.5 4.0 5.8 5.2 Annual Winter Spring Summer Autumn Snow (%) 55.3 73.6 65.2 51.8 Observed Freezing rain/ +0.14 +0.04 +0.37 +0.04 +0.11 0.4 0.3 0.2 1.0 (1912–2000) drizzle (%) Projected Hail, snow +0.61 +0.99 +0.52 +0.29 +0.62 2.4 1.8 3.2 2.4 (1961–2050) crystals (%)

Førland & Hanssen-Bauer 2003: Polar Research 22(2), 113–124 119 Fig. 4. Cumulative distribution (a) of air temperature during pre- cipitation as (a) snow, (b) sleet, (c, opposite page) rain and (d, opposite page) drizzle.

(b)

of reindeer was reported after severe crust and The low frequencies of freezing rain/drizzle ground ice formation in November and Decem- at Spitsbergen taken into consideration, it seems ber 1993. Table 5 shows that episodes with rain that it is rather seldom that events with compre- or drizzle falling at temperatures below 0 °C are hensive snow crust or ice formation are caused by rather infrequent at Spitsbergen; it occurs on liquid precipitation at temperatures below zero. average just once a year in Longyearbyen, Sval- Other causes include melting and refreezing at bard Airport and Ny-Ålesund. At Jan Mayen it is the snow surface, or rain absorbed and subse- more frequent, with nearly 10 cases per year. quently frozen in the surface snow layer.

Table 5. Threshold temperature for equal probability of Fractions of snow and rain as a function of air solid and liquid precipitation, and number of events (4 temperature observations/day) per year with liquid precipitation observed for T < 0 °C. Figure 5 demonstrates that there are distinct geographical differences in fractions of liquid/ Threshold temp. Station name Period Events/yr (°C ) solid precipitation as a function of air tempera- ture. For near zero temperatures on Bjørnøya and Bjørnøya 1956–1999 0.84 3.0 Jan Mayen, the fraction of liquid precipitation is Svalbard Airport 1975–1999 1.70 1.1 generally higher than at the Spitsbergen stations. Longyearbyen 1957–1977 1.96 1.0 On Bjørnøya, the precipitation is liquid in 90 % Ny-Ålesund 1975–1999 1.62 1.0 of precipitation events when the temperature is Jan Mayen 1956–1999 1.03 8.5 above 1.5 °C, while in Longyearbyen, Svalbard

120 Past and future climate variations in the Norwegian Arctic (c)

(d)

Airport and Ny-Ålesund this threshold is reached ground during summer; (2) during most of the at around 3 °C. year, Bjørnøya and Jan Mayen are farther from The “threshold” temperature—where the the sea ice border than the Spitsbergen stations, probability for liquid and solid precipitation and cold air masses from sea ice covered areas is equal— differs at the individual Arctic sta- are better mixed because of the longer travel dis- tions (Table 5). At the stations on Bjørnøya and tance over open sea. Jan Mayen, this threshold temperature is below or close to 1.0 °C, while at the Spitsbergen sites it is higher than 1.5 °C. The same pattern is also Trends in annual amounts of solid found for sleet (Fig. 4b), where the median tem- and liquid precipitation perature is lower on Bjørnøya and Jan Mayen than at the Spitsbergen stations. This indicates At Hopen and Svea, ca. 60 % of the annual precip- that, at the same 2 m air temperature, the air mass itation amount during the period 1975–2001 was aloft during precipitation is colder over Spitsber- reported as snow, and only about 20 % as rain. At gen than over Bjørnøya and Jan Mayen. Bear- Svalbard Airport and Ny-Ålesund, the fi gures ing in mind that rain or snow in this region may were ca. 45 and 25 %, while ca. 25 % fell as sleet occur throughout the year, two possible reasons or a mixture of rain and snow. On Bjørnøya and for this apparently more stably stratifi ed air over Jan Mayen, the amounts of snow, rain and mixed the island stations during precipitation might be: precipitation were nearly equal. These fractions (1) the contribution from convective precipitation are based on semi-daily measurements of precipi- over Spitsbergen caused by solar heating of the tation amounts and the precipitation types report-

Førland & Hanssen-Bauer 2003: Polar Research 22(2), 113–124 121 Fig. 5. Fractions of observations classifi ed as liquid precipitation at different temperatures (ra and sn are amounts of precipitation as rain and snow, respectively).

ed by the observers. No correction for gauge tion gauges. As the gauge undercatch is different undercatch has been performed and, according- for snow and rain, and further depends on wind ly, the fraction of solid precipitation is underes- and temperature, changes in climate will result in timated. changes in the undercatch of the gauges. Reduced The fractions of solid precipitation have dimin- fractions of annual precipitation falling as snow ished at all stations during the latest decades, lead to a reduced annual gauge undercatch, and particularly at Svalbard Airport and Jan Mayen. thus a fi ctitious positive trend for precipitation Based on linear regression, the fraction of solid even if the true precipitation does not change at precipitation on Jan Mayen is reduced from 39 % all (Førland & Hanssen-Bauer 2000). The poten- in 1975 to 20 % in 2001. The fraction of annual tial for such artifi cial trends is maximum in areas precipitation reported as mixed precipitation (i.e. with strong winds and where a large percentage sleet, or a combination of rain and snow during of the annual precipitation is solid, for example, the 12h sampling interval) has increased at all in the Norwegian Arctic. stations. By applying the rough correction factors pre- sented earlier on the annual amounts of solid, liquid and mixed precipitation it is possible to Fictitious trends in precipitation give crude estimates of “true” precipitation (Pc). amounts In Table 6, the same correction factors for solid, liquid and mixed precipitation have been applied Precipitation records from the Arctic are infl u- to all the Arctic stations, with no consideration enced by substantial measuring errors, e.g. taken to differences in wind and temperature caused by undercatch of conventional precipita- conditions at the stations. The measured values

Table 6. Changes in precipitation, 1975–2001. Pm is mean measured annual precipitation. Mean correction factor is the ratio between corrected (Pc) and measured (Pm) precipitation. Change is the difference between levels in 1975 and 2001, based on linear regression. Change for measured and corrected precipitation is given both in millimetres and as a percentage per decade of the 1975 level.

Precipitation change

Pm Mean correction Pc Measured Corrected (mm) factor (mm) Change (mm) Change (%) Change (mm) Change (%) Bjørnøya 396 1.52 602 129 15.0 198 15.2 Hopen 469 1.64 767 –65 –5.0 –119 –5.6 Sveagruva 271 1.66 451 –40 –5.2 –100 –7.6 Svalbard Airport 192 1.56 299 10 2.1 5 0.7 Ny-Ålesund 403 1.56 631 88 9.4 142 9.8 Jan Mayen 680 1.48 1007 –97 –5.2 –203 –7.2

122 Past and future climate variations in the Norwegian Arctic are not corrected for evaporation and wetting trend in measured precipitation. effects. The rough estimates of Pc presented in There are substantial differences between Table 6 indicate that the true precipitation at all measured and “true” precipitation in the Arctic, stations in the Svalbard region is more than 50 % both with respect to material amounts as well higher than the measured values. as trends. Accordingly, corrected precipitation The estimates in Table 6 are based on a short values should be used in studies of historical time period and rather crude approximations, but trends as well as for monitoring future trends. nonetheless illustrate the importance of correct- Estimates of “true” precipitation are also impor- ing Arctic precipitation series for undercatch. For tant for water balance assessments in the Arctic. most stations there are substantial differences between trends based on measured and corrected values, both in material amounts and also for rel- Acknowledgements.—Thanks to Jan-Gunnar Winther at the ative changes. For instance, the increase in meas- Norwegian Polar Institute for useful suggestions, and to two ured annual precipitation at Svalbard Airport anonymous reviewers who contributed to the improvement of during the period 1975–2001 is 2.1 % per decade. the manuscript. However, by correcting for gauge undercatch, the increase in the resulting estimates for “true” pre- cipitation is 0.7 % per decade. The scenarios cur- References rently available (Table 3) indicate a 3 °C increase in annual temperature from the present to 2050 Aagaard, K. & Carmack, E. C. 1989: The role of sea ice and in Svalbard. For an increase in annual tempera- other fresh waters in the Arctic circulation. J. Geophys. ture of 4 °C, Førland & Hanssen-Bauer (2000) Res. 94(C10), 1485–1498. estimated a fi ctitious precipitation increase of Benestad, R., Førland, E. J. & Hanssen-Bauer, I. 2002: Empirically downscaled temperature scenarios for Sval- 2 % per decade. This virtual increase, which is bard. Atmos. Sci. Lett., doi:10.1006/Asle.2002.0051. caused solely by reduced measuring errors, is Birkeland, B. J. 1930: Temperaturvariationen auf Spitz- thus of the same magnitude as the projected pre- bergen. (Temperature variations on Spitsbergen.) Meteor- cipitation increase of 1.4 % per decade (Table ol. Z. 47, 234–236. 2) under global warming. This virtual increase Bjørge, D., Haugen, J. E. & Nordeng, T. E. 2000: Future climate in Norway. Dynamical experiments within the will be measured in addition to an eventual real RegClim project. Res. Rep. 103. Oslo: Norwegian Mete- increase. orological Institute. Dai, A., Fung, I. Y. & Del Genio, A. D. 1997: Surface observed land precipitation variations during 1900–88. J. Conclusions Clim. 10, 2943–2962. Folland, C. & Karl, T. R. 2001: Observed climate variabili- ty and change. In International Panel on Climate Change: Annual temperatures in the Longyearbyen/Sval- Climate change 2001: the scientifi c basis. Contribution of bard Airport area have increased by more than Working Group I to the third assessment report of the Inter- 1 °C since 1910, but because of large interannu- governmental Panel on Climate Change. Pp. 99–181. Cam- bridge: Cambridge University Press. al and decadal variations this trend is not statis- Førland, E. J., Allerup, P., Dahlström, B., Elomaa, E., Jóns- tically signifi cant. Although the temperature has son, T., Madsen, H., Perälä, J., Rissanen, P., Vedin, H. increased signifi cantly since the cold 1960s, the & Vejen, F. 1996: Manual for operational correction of present temperature level is still lower than in the Nordic precipitation data. Klima 24/96. Oslo: Norwegian Meteorological Institute. 1930s at all stations in the Norwegian Arctic. Førland, E. J. & Hanssen-Bauer, I. 2000: Increased precipita- Measured annual precipitation in the Svalbard tion in the Norwegian Arctic: true or false? Clim. Change region and at Jan Mayen has increased substan- 46, 485–509. tially (15 - 25 %) during the last 80 to 90 years. At Førland, E. J., Hanssen-Bauer, I. Jónsson, T., Kern-Hansen, Bjørnøya, Svalbard Airport and Jan Mayen, the C., Nordli, P. Ø., Tveito, O. E. & Vaarby Laursen, E. 2002: Twentieth century variations in temperature and precipita- positive precipitation trend during the 20th cen- tion in the Nordic Arctic. Polar Rec. 38, 203–210. tury is statistically signifi cant. Førland, E. J., Hanssen-Bauer, I. & Nordli, P. Ø. 1997a: Cli- The fraction of annual precipitation falling as mate statistics and long-term series of temperature and snow has decreased at all stations in the Norwe- precipitation at Svalbard and Jan Mayen. Klima 21/97. Oslo: Norwegian Meteorological Institute. gian Arctic during recent decades. This leads Førland, E. J., Hanssen-Bauer, I. & Nordli, P. Ø. 1997b: Oro- to a reduced annual undercatch in the precipita- graphic precipitation at the glacier Austre Brøggerbreen, tion gauges, and consequently a fi ctitious positive Svalbard. Klima 02/97. Oslo: Norwegian Meteorological

Førland & Hanssen-Bauer 2003: Polar Research 22(2), 113–124 123 Institute. of Canadian daily precipitation time series. Atmos.–Ocean Fu, C., Diaz, H. F., Dong, D. & Fletcher, O. 1999: Changes 37, 53–85. in atmospheric circulation over the Northern Hemisphere Mysak, L., Manak D. K. & Marsden, R. F. 1990: Sea ice oceans associated with the rapid warming of the 1920s. Int. anomalies observed in the Greenland and Labrador seas J. Climatol. 19, 581–606. during 1901–1984 and their relation to an interdecadal Goodison, B. E., Louie, P. Y. T. & Yang, D. 1998: WMO solid climate cycle. Clim. Dyn. 5, 111–133. precipitation measurement intercomparison. WMO/TD- Nordli, P. Ø., Hanssen-Bauer, I. & Førland, E. J. 1996: Homo- Rep. 872. Geneva: World Meteorological Organization. geneity Analyses of temperature and precipitation series Hanssen-Bauer, I. 2002: Temperature and precipitation in from Svalbard and Jan Mayen. Klima 16/96. Oslo: Norwe- Svalbard 1912–2050: measurements and scenarios. Polar gian Meteorological Institute. Rec. 38, 225–232. Parker, D. E. & Alexander, L. 2002: Global and regional cli- Hanssen-Bauer, I & Førland, E. J. 1994: Homogenizing long mate in 2001. Weather 57, 328–340.

Norwegian precipitation series. J. Clim. 7, 1001–1013. Räisänen, J. 2001: CO2 induced climate change in CMIP2 Hanssen-Bauer, I. & Førland, E. J. 1998a: Long-term trends experiments: quantifi cation of agreement and role of inter- in precipitation and temperature in the Norwegian Arctic: nal variability. J. Clim. 14, 2088–2104. can they be explained by changes in the atmospheric circu- Roeckner, E., Bengtsson, L., Feichter, J., Lelieveld, J. & lation patterns? Clim. Res. 10, 143–153. Rodhe, H. 1999: Transient climate change simulations with Hanssen-Bauer, I. & Førland, E. J. 1998b: Annual and seasonal a coupled atmosphere–ocean GCM including the tropo- precipitation variations in Norway 1896–1997. Klima 27/98. spheric sulphur cycle. J. Clim. 12, 3004–3032. Oslo: Norwegian Meteorological Institute. Sand, K., Winther, J.-G., Maréchal, D., Bruland, O. & Mel- Hanssen-Bauer, I. & Førland, E. J. 2001: Verifi cation and vold, K. 2003: Reginal variations of snow accumulation on analysis of a climate simulation of temperature and pres- Spitsbergen, Svalbard, 1997–99. Nord. Hydrol. 34, 17–32. sure fi elds over Norway and Svalbard. Clim. Res. 16, 225– Serreze, M. C., Walsh, J. E., Chapin III, F. S., Osterkamp, T., 235. Dyurgerov, M., Romanovsky, V., Oechel, W. C., Morison, Hanssen-Bauer, I., Solås, M. K. & Steffensen, E. L. 1990: J., Zhang, T. & Barry, R. G. 2000: Observational evidence The climate of Spitsbergen. Klima 39/90. Oslo: Norwegian of recent change in the northern high-latitude environment. Meteorological Institute. Clim. Change 46, 159–207. Hesselberg, T. & Johannessen, T. W. 1958: The recent vari- Steffensen, E. 1969: The climate and its recent variations at ations of the climate at the Norwegian Arctic stations. In the Norwegian Arctic stations. Meteorol. Ann 5, 217–349. R. C. Sutcliffe (ed): Polar Atmosphere Symposium, part I, Steffensen, E. 1982: The climate at Norwegian Arctic sta- meteorological section, Oslo, Norway, 2–8 July, 1956, Pp. tions. Klima 5, 3–44. Oslo: Norwegian Meteorological 18–29. London: Pergamon Press. Insti tute. Hisdal, V. 1976: . A short survey. Vinje, T. 1982: Frequency distribution of sea ice in the Green- Tromsø: Norwegian Polar Institute. land and Barents seas, 1971–80. Nor. Polarinst. Årb. 1980, Houghton J. T., Callander, B. A. & Varney, S. K. 1992: Cli- 57–61. mate change 1992. Cambridge: University Press. Walsh, J. E., Kattsov, V., Portis, D. & Meleshko, V. 1998: Hulme, M. 1995: Estimating global changes in precipitation. Arctic precipitation and evaporation: model results and Weather 50, 34–42. observational estimates. J. Clim. 11, 72–87. IPCC (International Panel on Climate Change) 2001: Climate Winther, J.-G., Bruland, O., Sand, K., Killingtveit, Å. & change 2001: the scientifi c basis. Contribution of Working Marechal, D. 1998: Snow accumulation distribution on Group I to the third assessment report of the Intergovern- Spitsbergen, Svalbard, in 1997. Polar Res. 17, 155–164. mental Panel on Climate Change. Cambridge: Cambridge Winther, J-G., Bruland, O., Sand, K., Gerland, S., Mare- University Press. chal, D., Ivanov, B., Głowacki, P. & König, M. 2003: Snow Mekis, E. & Hogg, W. D. 1999: Rehabilitation and analysis research in Svalbard. Polar Res. 22, 125–144 (this issue).

124 Past and future climate variations in the Norwegian Arctic